Complete genome assembly of clinical multidrug resistant Bacteroides fragilis isolates enables comprehensive identification of antimicrobial resistance genes and plasmids.
This repository supports our paper, which is currenctly available as a preprint from BioRXiv:
Content includes:
- Code and commands for calling community-built and published tools
- Links to data repositories, including raw and processed reads (fast5 and fastq) as well as genome assemblies
- Supplementary results and links to supplementary files
Bacteroides fragilis is an opportunistic pathogen that resides in the human gut as part of the human commensal flora. Short DNA sequencing alone is unable to resolve the structure of the genome. VSeven complete genome assemblies have been deposited to public databases to date (2019-02-2018). To add to the public databases and to explore the location of antimicrobial resistance genes in the context of the complex genomes of B. fragilis, we aimed to complete genome assembly using a hybrid assembly approach. Using the Sanger sequenced Bacteroides fragilis NCTC 9343(NCBI RefSeq accession GCF_000025985.1) as a comparative reference, we sequenced B. fragilis CCUG4856T (=NCTC9343=ATCC25285) with Illumina 150 bp PE and Oxford Nanopore (ONT) MinION Rapid barcoding kit and assembled using various assemblers and polishing tools. The best assembly was with hybrid Unicycler with FilterByTile and Cutadapt filteres Illumina reads and FiltLong filtered and Canu corrected ONT reads. This approach was then applied to six multidrug resistant isolates. Resistance genes and IS elements were identified. Plasmids where charaterised.
Manuscript available as pre-print on bioRxiv doi: 10.1101/633602
filterbytile from BBMap
bbmap/36.49
filterbytile.sh in1=<short reads 1> in2=<short reads 2> out1=filtered1.fq.gz out2=filtered2.fq.gz
trim_galore/0.4.4
trim_galore --paired --quality 10 <input 1> <input 2>
seqtk/1.0-r82-dirty pigz/2.3.4
#set estimated genome length
GLEN=5.3
#set target max depth
DEPTH=100
BASES=$(zcat val_1.fq.gz | seqtk seq -A | grep -v "^>" | tr -dc "ACTGNactgn" | wc -m)
BASES=$(expr $BASES \* 2) #total bases is double R1. need to escape the asterix
ORI_DEPTH=$(expr $BASES / $GLEN) #calculate original (crude) depth
echo DEPTH: $DEPTH
calc() { awk "BEGIN{print $*}"; } #need this short awk funktion as bash expr only calculates integer
FACTOR=$(calc $DEPTH / $ORI_DEPTH)
if [ "$ORI_DEPTH" -gt "$DEPTH" ]; then #subsample
seqtk sample -s100 val_1.fq.gz $FACTOR | pigz --fast -c -p $NPROCS > R1.sub.fq.gz
seqtk sample -s100 val_2.fq.gz $FACTOR | pigz --fast -c -p $NPROCS > R2.sub.fq.gz
else
fi
Nanopore reads were demultiplexed following the notes by Ryan Wick. Raw reads were demultiplexed with Deepbinner and then basescalled with Albacore, retaining only reads where the two agreed on the barcode. PoreChop was then used to trim adapters and barcodes from reads.
Running PoreChop
porechop_out="$b"_trimmed.fastq.gz
porechop_log="$b"_porechop.log
porechop --threads $NPROCS \
--check_reads 100 \
--discard_middle \
-i demultiplexed_fastqs/"$b"_untrimmed.fastq.gz \
2> "$porechop_log" | pigz -p $NPROCS > "$porechop_out"
inp1=input1.fastq.gz
inp2=input2.fastq.gz
NPROCS=<number of processes/threads>
output=output.fastq.gz
zcat $inp1 $inp2 | pigz --fast -c -p $NPROCS > $output
Filtlong 0.2.0
filtlong --min_length 100 --keep_percent 90 --target_bases 500000000 input.fastq.gz | gzip > output.fastq.gz
Canu correction of ONT reads
Canu 1.8 run with standard options or corOutCoverage=999 or corMinCoverage=0
canuprefix=$(basename output.fastq.gz .fastq.gz)
canu -correct \
-p $canuprefix \
-d <outdir> \
genomeSize=5.4m \
-nanopore-raw output.fastq.gz \
useGrid=false
Canu assembly
canu \
-p canu.ont \
-d canu_ont \
genomeSize=5.2m \
useGrid=false \
-nanopore-raw <path to nanopore reads>
maxMemory=<set memory> \
maxThreads=<set threads>
Unicycler hybrid assembly
Unicycler 0.4.7 Dependencies: jre/1.7.0 pilon/1.22 racon/1.3.1 spades/3.13.0 samtools/1.9
unicycler \
--short1 <path to short reads 1> \
--short2 <path to short reads 2> \
--long <path to long reads> \
--out <output path> \
--threads=<number of threads> \
--verbosity=2 \
--min_fasta_length=100 \
--mode normal
Flye ONT read assembly
Flye 2.3.7
flye --nano-raw <long reads> \
--genome-size 5.3m \
--threads <number of threads> \
--out-dir <outdir>
Flye ONT read assembly with Flyepolish
Flye 2.3.7 run with three iterations of Flyes internal polishing engine
flye --nano-raw <long reads> \
--genome-size 5.3m \
--threads <number of threads> \
--out-dir <outdir> \
-i 3
miniasm assembly
minimap2/2.6 miniasm/0.3r179
minimap2 -x ava-ont -t $NPROCS ont_reads.fastq ont_reads.fastq | gzip -1 \
> output.reads.paf.gz
miniasm -f <ONT reads> output.reads.paf.gz > output.miniasm.gfa
# convert to fasta after lh3 https://www.biostars.org/p/169516/
awk '/^S/{print">"$2"\n"$3}' output.miniasm.gfa | \
fold > output.miniasm.fasta
Skesa assembly
skesa/2.3.0
skesa --fastq <short reads 1> \
--fastq <short reads 2> \
--cores <number of cores> \
--memory <allocated memory> \
--contigs_out output.skesa.fasta
SPAdes assembly
spades/3.13.0
spades.py --threads <number of cores> \
--memory <allocated memory> \
-k 31,51,71 \
--pe1-1 <short reads 1> \
--pe1-2 <short reads 2> \
-o output.spades \
--careful \
--tmp-dir <scratch directory>
HybridSPAdes assembly
spades/3.13.0
spades.py --threads <number of cores> \
--memory <allocated memory> \
-k 31,51,71 \
--pe1-1 <short reads 1> \
--pe1-2 <short reads 2> \
--nanopore <nanopore reads> \
-o output.hybridspades \
--careful \
--tmp-dir <scratch directory>
Assemblies were polished with Nanopolish, Racon, and/or Pilon. nanopolish_runner.sh, ../master/pilon_runner_nochange.sh, racon_runner.sh, and ../master/racon_illumina_runner.sh were heavily inspired by @nataliering 's scripts.
Usage: Nanopolish_runner.sh
run_nanopolish <directory_of_fast5s> <path/to/sequencing_summary.txt> <reads_from_albacore.fastq> <draft_assembly.fasta> <output_name.fasta>
pilon_runner_nochange.sh
pilon_runner <in.fasta> <illumina_S1> <illumina_S2> <out_prefix>
racon_runner.sh
racon_runner <assembly.fasta> <reads.fastq> <out.fasta>
racon_illumina_runner.sh
racon_illumina_runner <assembly.fasta> <illuminareads1.fastq.gz> <reads2.fastq.gz> <out.fasta>
Assemblies in .fasta format were collected in the same folder, and the following scripts run in that folder. This produced a number of output directories and .tsv files, with concatenated results for all .fasta files. These .tsv files could then be merged, creating the supplementary file XXX with evaluation results.
The following runs ALE for all *fasta files in a folder, and extracts the score from the individual ALE results files and writes the scores to a seperate .tsv file.
mkdir ale
#create file for aggregation of results
printf "Assembly\total_ALE_score\n" > ale/totalALE.tsv
#analyse all fasta
for F in *.fasta; do
N=$(basename $F .fasta )
bwa index $F
bwa mem -t $NPROCS $F $S1 $S2 \
| samtools sort -@ $NPROCS > $sp/$N.ale.aln.bam
ALE --nout $sp/$N.ale.aln.bam $F ale/$N.output.ale.meta.txt
#cleanup
rm $sp/$N.ale.aln.bam
rm $F.amb $F.ann $F.bwt $F.pac $F.sa
rm ale/$N.output.ale.meta.txt.param
#extract info and print in results file
totalALE=$(head -n 1 ale/$N.output.ale.meta.txt | tr ' ' \\t | cut -f3)
printf "$F\t$totalALE\n" >> ale/totalALE.tsv
done
Average nucleotide identity was calculated using @chjp's ANI.pl
Reference genome was RefSeq GCF_000025985.1.
ref_genome=GCF_000025985.1_ASM2598v1_genomic.fna
allfasta=$(ls *.fasta)
printf "Assembly\tANI_pl\n" > ani.pl.results.txt
for F in $allfasta; do
N=$(basename $F .fasta | sed s/\[.]/_/g)
assembly=$(basename $F .fasta)
perl ANI.pl \
-bl <path to blastall> \
-fd <path to formatdb> \
-qr $F \
-sb $ref_genome \
-od "$N"_"ani" \
>> $assembly.ani.pl.txt
#concatenate ani.pl results
printf "$assembly\t" >> ani.pl.results.txt
grep -w -F -e ANI $assembly.ani.pl.txt | sed 's/.*\ //' >> "ani.pl.results.txt"
#printf "\n" >> "ani.pl.results.txt" #need newline after
#cleaning
rm -r "$N"_"ani"
rm $assembly.ani.pl.txt
done
allfasta=$(ls *.fasta)
mkdir busco_out
cd busco_out
for F in $allfasta; do
echo "BUSCO file $F"
N=$(basename $F .fasta | sed s/\[.]/_/g)
python run_BUSCO.py \
--in ../$F \
--out $N \
--lineage_path <path to bacteroidetes_odb9> \
--mode genome \
--cpu $NPROCS \
--tmp_path $sp/busco_tmp \
--force
#cleaning
rm -r run_$N/augustus_output
rm -r run_$N/blast_output
rm -r run_$N/checkpoint.tmp
rm -r run_$N/hmmer_output
done
binfolder="./"
outputfolder=checkm_out
markerfile=checkmmarkerfile
checkm tree $binfolder $outputfolder -x fasta --threads $NPROCS
checkm tree_qa $outputfolder
checkm lineage_set $outputfolder $markerfile
checkm analyze $markerfile $binfolder $outputfolder -x fasta --threads $NPROCS
checkm qa $markerfile $outputfolder --out_format 2 \
--threads $NPROCS --file $outputfolder/checkm_qa_out.tsv --tab_table
#cleaning
rm -r checkm_out/bins
rm -r checkm_out/storage
dnadiff from MUMmer
for F in *.fasta; do
N=$(basename $F .fasta | sed s/\[.]/_/g)
dnadiff --prefix dnadiff/$N.dnadiff $ref_genome $F
#get 1-to-1 AvgIdentity for QRY and print to file
avgid_out=dnadiff/mummerdnadiff_results.txt
printf "$F""\t">> $avgid_out
grep -w -F -e AvgIdentity dnadiff/$N.dnadiff.report | \
head -n 1 | tail -c 7 | tr -d '[:space:]' >> "$avgid_out"
printf "\n" >> "$avgid_out" #need newline after
#cleaning
rm dnadiff/$N.dnadiff*
done
Prokka annotation
Prokka is run for all assemblies in .fasta format. The reference proteins of GCF_000025985.1 is used as protein database for prokka. Counts for CDS, gene, RNA ect is then collected into one .tsv file.
ref_features=GCF_000025985.1_ASM2598v1_genomic.gbff
allfasta=$(ls *.fasta)
for F in $allfasta; do
echo "PROKKA file $F"
N=$(basename $F .fasta | sed s/\[.]/_/g)
prokprefix=$(basename $F .fasta)
prokka --compliant \
--outdir prokka/$N --force \
--prefix $prokprefix \
--proteins $ref_genomic_gbff \
--cpus $NPROCS \
$F
#cleaning
rm prokka/$N/*.err
rm prokka/$N/*.ffn
rm prokka/$N/*.sqn
rm prokka/$N/*.fsa
done
#Gather results in prokkaresults.tsv
#prepare column headers for prokkaresults.tsv
printf "Prokka bases\tProkka CDS\tProkka contigs\tProkka CRISPR\tProkka genes\tProkka isolate\tProkka rRNA\tProkka tmRNA\tProkka tRNA\n" > prokkaresults.tsv
for F in */*.txt; do
F_base=$(basename $F .txt)
# "isolate" and the basename to first line in tempfile.tsv
printf "isolate\t$F_base\n" > tempfile.tsv
#remove the first line of the prokka result.txt, convert space to tab, remove the colon
tail -n +2 $F | tr ' ' \\t | tr -d : >> tempfile.tsv
# we have to sort alfabetically as prokka does not output results to the txt in the same order for
# each annotation run.
# sort the tempfile cut the second column and paste as a row (transpose), and add this to prokkaresults.tsv
sort tempfile.tsv | cut -f2 | paste -s >> prokkaresults.tsv
done
#Reorder columns for more logical output.
awk 'BEGIN {FS="\t"}{OFS="\t"}{print $6,$2,$5,$1,$3,$7,$8,$9,$4}' prokkaresults.tsv > prokkaresults_reorder.tsv
Several statistics were collected from Quast GCF_000025985.1 were used as reference for Quast.
allfasta=$(ls *.fasta)
ref_genome=GCF_000025985.1_ASM2598v1_genomic.fna
ref_features=GCF_000025985.1_ASM2598v1_genomic.gff
quast.py $allfasta \
-r $ref_genome \
--features $ref_features \
-o quast_output \
--threads $NPROCS \
--min-contig 500 \
--k-mer-stats \
--plots-format png \
--no-icarus \
--no-html
#cleaning
cd quast_output
rm -r aligned_stats
rm -r k_mer_stats
rm -r basic_stats
rm -r contigs_reports
rm -r genome_stats
Identification of resistance genes and IS elements with ABRicate
ABRicate does not output the orientation of a hit in the results. To ascertain if an IS element is oriented upstream and in the opposite strand of the resistance gene of interest, a small change was made to the ABRicate script prior to use. See ABRicate issue 83. Resistance genes were identified using the Resfinder, CARD and NCBI databases that come with ABRicate. Gene sequences of the putative multidrug efflux pumps BexA (GenBank: AB067769.1:3564..4895) and BexB (GenBank: AY375536.1:4599..5963) were downloaded from the NCBI Nucleotide database. The ISFinder Insertion Sequence database was downloaded from here and duplicates removed (see this issue. The above databases and sequences were collated into one ABRicate database, abricate_tvs, following the guide on ABRicates github site. ABRicate was run with very low --minid and --mincov settings, as we previously had experienced missing hits (from split genes or low homology sequences).
abricate --db abricate_tvs --minid 40 --mincov 25 --threads 3 <contigs.fasta> > out.tab
We wanted further information and evidence supporting whether circularised contigs were indeed plasmid sequences. Increased relative coverage compared to the main replicon/chromosome is indicative of plasmid sequence, as is circularisation. Sequences of the plasmid replication domain families, listed in Table 1 in Jørgensen et al 2014 were downloaded from the Pfam database. Specifically the full-length sequences for all sequences in the full alignments were downloaded for: PF01051 PF01402 PF01446 PF01719 PF01815 PF02486 PF03090 PF03428 PF04796 PF05732 PF06504 PF06970 PF07042 PF10134
The following awk one liner was used to improve the sequence names for use with ABRicata. The example below is for PF03428_full_length_sequences.fasta
awk -F '>' '/^>/ { $1 = ">pfam~~~RP-C~~~" } { gsub (/[()]/,"",$2) } { gsub (" ","\t",$2) } { gsub ("\t"," PF03428.8 RP-C Replication protein C N-terminal domain 0 ",$2) }{ print $1 $2}' PF03428_full_length_sequences.fasta > PF03428_RP-C_abr.fasta
The individual *_abr.fasta files were concatenated, but the concatenated file contained duplicates. fasta_unique by @b-brankovics was used to remove duplicates. The removed duplicates were writen to a .tab file for later review, and Pfam_replicationdomain_unique.fasta used to build a protein sequence database with ABRicate.
cat *_abr.fasta > Pfam_replicondomains.fasta
perl fasta_unique.perl Pfam_replicationdomain.fasta> Pfam_replicationdomain_unique.fasta 2>Pfam_replicationdomain_unique.tab
# Local BLAST of np reads to assembly
'Filter low Complexity'
Expect 0,00000000000000000000000000000001
Word size 11
Match/mismatch 'Match 1, Mismatch -2'
Gap costs Existence 1, Extension 1
Max number of hit sequences 500
The NCBI/ENA Bioprojects contain sequence data and final assemblies:
| Strain | Alternative desigatores | Bioproject accession |
|---|---|---|
| CCUG4856T | ATCC25285, NCTC9343 | PRJNA525024 |
| BF017 | O17, BF17, DCMOUH0017B | PRJNA244943 |
| BFO18 | O18, BF18, DCMOUH0018B | PRJNA244944 |
| S01 | DCMSKEJBY001B | PRJNA244942 |
| BFO42 | O42, BF42, DCMOUH0042B | PRJNA253771 |
| BFO67 | O67, BF67, DCMOUH0067B | PRJNA254401 |
| BFO85 | O85, BF85, DCMOUH0085B | PRJNA254455 |
ONT data in fastq format, demultiplexed and with barcodes and adapters removed can be found here:
The 141 variations of assemblies and polished assembleis of CCUG4856T are available at
The stages of genome assemblies for the complete circular assemblies of the seven isolates are available at
